OpenAI's GPT-Astra, GPT-6.1 Sol and GPT-6 Luna reach Google Cloud's Model Garden, but only as an invite-only preview with general availability weeks away
Three OpenAI models, one Google Cloud invoice, and spend that draws down existing commitments. Verdict: platform teams should request preview access now, but nothing here is production-ready and no pricing or benchmarks were published.
Google Cloud customers will be able to purchase OpenAI frontier models via model garden, transacted through the Google Cloud Marketplace. Enterprises with Google Cloud spend commitments will be able to draw down their GCP commitments for these purchases. General availability is planned in the coming weeks. Microsoft remains OpenAI's primary cloud partner.
Comprehensive take
This is a distribution story, not a capability story. The meaningful signal is that OpenAI models now sit on all three of the largest clouds, which makes the model a swappable line item and moves competition toward price, latency and governance. Until prices and a general-availability date appear, request preview access and keep your evaluation portable.
Use cases
Teams with Google Cloud commitments
Marketplace purchase lets OpenAI usage count against existing spend, per Google's announcement.
Regulated-industry platform teams
Google positions its access controls, data governance and isolated gateways as the control layer around the models; verify them in your own review.
Codex users
Google says local Codex CLI and Desktop workflows can already be pointed at its platform for inference in the preview.
Specs
| Models named | GPT-Astra, GPT-6.1 Sol, GPT-6 Luna source |
|---|---|
| Status (Google) | Private preview to select enterprise customers, with more models coming soon source |
| Status (OpenAI) | General availability planned in the coming weeks; invite-only preview for development and internal use, not production source |
| Billing | Google Cloud Marketplace, single invoice, draws down existing commitments source |
| Confidential Compute | Listed as coming soon source |
| Distribution context | Exclusive Microsoft arrangement expired in April; AWS added earlier source |
Insights
- Axion reading: the two sources disagree on timing (preview now versus general availability in coming weeks); we treat the more conservative OpenAI statement as the planning date.
- Axion note: no price, latency or benchmark figures were published for these models on Google Cloud, so any cost comparison with Gemini or Claude is not yet possible.
What was announced
OpenAI and Google Cloud said on October 21 that OpenAI's frontier models and its Codex agent workflows are coming to Google Cloud's Model Garden. Google's own announcement, written by Michael Gerstenhaber, names three models: GPT-Astra, aimed at coding, computer use, professional tasks and scientific research; GPT-6.1 Sol, described as near-Astra intelligence at lower cost; and GPT-6 Luna, a fast, cost-efficient model for everyday work at scale. Reuters reports the flagship as "GPT-6 Astra", so expect naming to settle once documentation catches up.
Preview now, general availability later
The two companies describe the timing differently, and the difference matters. Google says the three models are rolling out now in private preview to select enterprise customers, and that developers can already connect a local Codex CLI or Desktop setup to Google's platform for inference. OpenAI's post says general availability is planned in the coming weeks, and its access form describes an invite-only preview limited to approved development and internal-testing use cases, not production. If you are planning a launch around these models on Google Cloud, treat the date as unannounced.
Why procurement teams will care
The practical change is commercial rather than technical. Purchases go through Google Cloud Marketplace, appear on a single Google Cloud invoice, and draw down existing cloud commitments. Google adds that usage counts toward its Flexible Savings Plans across model providers. For an organisation that has already committed spend to Google Cloud, that removes a separate vendor contract and a separate security review of a new data path. Google also says queries are routed through isolated gateways on dedicated infrastructure, with Confidential Compute listed as coming soon.
What is missing
Neither company published prices, rate limits, context windows, regional availability or benchmark results for the three models on this platform. Google's claim that it is now the only hyperscaler offering all three leading closed-source LLMs is its own framing and is not independently verified here. OpenAI also states that Microsoft remains its primary cloud partner, and Reuters notes the exclusive Microsoft distribution arrangement expired in April, with Amazon Web Services already added.
Who should act
If you already run workloads on Google Cloud and want to compare OpenAI models against Gemini and Claude behind one API and one bill, file for preview access and design your evaluation harness so the model is a configuration value. If you are not a Google Cloud customer, this changes little today: the models are the same ones available elsewhere, and the new element is the buying channel.